No middleman. No account. Your data stays on the phone.
The app talks straight from your iPhone to the AI provider you choose, using your own key. This page shows exactly what stays on the device, the one thing that crosses to your provider, and the parts we are honest about not doing.
Two places see your chats. Almost everything lives in one.
There is your phone, and there is the AI provider you picked. Nearly everything the app touches stays on the phone. Exactly one thing crosses over to your provider, and only when you send a message or a brief you set up runs.
- Your chats and history
- Everything your agent remembers
- Your API keys iOS Keychain
- Your settings and personalization
Anthropic, OpenAI, Gemini, or another. It receives the masked request, authenticated with your key, and sends an answer back. The same as using their own app with your own key. We are not in the path.
The whole journey of a single request.
Say you tell your agent text Cass at 2pm. Here is everything that happens to it.
You type it
The words sit in the on-device database on your iPhone. Nothing has left yet.
It is masked on the device
Before anything is sent, the app swaps out names, emails, and phone numbers for placeholders. Cass never leaves the phone as a name.
leaves as → text {{PERSON_0}} at 2pmIt goes straight to your provider
The masked request travels from your phone to the provider you chose, signed with your key from the Keychain. There is no PrentusAI server in between.
The reply is restored on the device
The answer comes back and the placeholder turns into Cass again, on your phone. You read a normal message. The provider only ever saw a token.
{{PERSON_0}} becomes → CassThe parts we would rather tell you.
Privacy claims are easy to make and easy to overstate. Here is where the line actually is.
Inference is not local
The model runs on your chosen provider's servers, not on your phone. If you need fully offline, on-device inference, the app is not that. What stays local is your data, your memory, and your keys.
Free-typed details are not masked
The masking covers names, emails, and phone numbers. If you type a street address or an account number straight into the chat, it reaches your provider as written. Treat the chat the way you would treat their own app.
It is closed source today
You cannot read the code yet, which is exactly why this page spells out what happens. We are open to an outside review of the data handling.
Wake-ups use one small service
If you turn on "Get briefs ready early", the app sends a small PrentusAI service a random ID, your iPhone's push address, and the times of your briefs, so it can wake the app before each one. It never receives a brief's question or answer, or anything from your chats, and it deletes what it holds when you turn the switch off. It is off unless you turn it on, and no version released so far, including 1.5.0, has the switch.
The shape of it, in plain terms.
Your chats never reach us.
- × No PrentusAI account or login
- × No PrentusAI server in the request path
- × No analytics, tracking, or ad SDKs in the app
Actions wait. Keys are yours.
Calendar events, messages, and reminders show you a card and do nothing until you tap to confirm. Text that your agent reads from a web page cannot act on its own.
And the app is free. You pay your provider directly for what you use. We do not resell tokens or take a cut.